def compute_attributes(self):
input_stream = cv.VideoCapture(self.file_name)
self.sample_rate = input_stream.get(cv.CAP_PROP_FPS)
input_stream.release()
# Use multi-threading by Adrian of https://pyimagesearch.com
input_stream = FileVideoStream(self.file_name).start()
directograms = []
directogram_times = []
frame = input_stream.read()
last_frame = cv.cvtColor(frame, cv.COLOR_BGR2GRAY)
self.shape = frame.shape
while input_stream.more():
if len(directograms) % 250 == 0:
print('Video Processing: {:.2f}s'.format(
len(directograms) / self.sample_rate))
frame = input_stream.read()
if not input_stream.more():
break
next_frame = cv.cvtColor(frame, cv.COLOR_BGR2GRAY)
# These parameters were taken from a demo in the documentation
flow = cv.calcOpticalFlowFarneback(last_frame, next_frame, None,
0.5, 3, 15, 3, 5, 1.2, 0)
directograms.append(get_directogram(flow))
directogram_times.append(len(directograms) / self.sample_rate)
last_frame = next_frame
self.envelope_times = np.array(directogram_times)
self.envelopes = get_impact_envelopes(directograms,
self.envelope_times)
self.envelope_sample_rate = self.sample_rate
input_stream.stop()
def fry_gif(update, url, number_of_cycles, args):
number_of_emojis = 1.5 if args['high-fat'] else 1 if args['low-fat'] else 0
bulge_probability = 0.3 if args['heavy'] else 0.15 if args['light'] else 0
magnitude = 4 if args['deep'] else 1 if args['shallow'] else 2
name = update.message.from_user.first_name
filename = '%s_%s_%s' % (update.message.chat_id, name,
update.message.message_id)
filepath = bin_path + '/temp/' + filename
caption = __get_caption(name, number_of_cycles, args)
output = bin_path + '/temp/out_' + filename + '.mp4'
gifbio = BytesIO()
gifbio.name = filename + '.gif'
fs = [__posterize, __sharpen, __increase_contrast, __colorize]
shuffle(fs)
if not __download_gif(url, filepath):
return
fvs = FileVideoStream(filepath + '.mp4').start()
frame = fvs.read()
height, width, _ = frame.shape
try:
fps = fvs.get(CAP_PROP_FPS)
except:
fps = 30
out = VideoWriter(output, VideoWriter_fourcc(*'mp4v'), fps,
(width, height))
out.write(
fry_frame(frame, number_of_cycles, fs, number_of_emojis,
bulge_probability, magnitude, args))
while fvs.more() or fvs.more():
try:
temp = fry_frame(fvs.read(), number_of_cycles, fs,
number_of_emojis, bulge_probability, magnitude,
args)
except Exception as e:
break
out.write(temp)
fvs.stop()
fvs.stream.release()
out.release()
update.message.reply_animation(open(output, 'rb'),
caption=caption,
quote=True)
try:
__upload_to_imgur(output, caption)
except (Exception, BaseException) as e:
print(e)
remove(filepath + '.mp4')
def run_video():
PATH_VIDEO = 'D:/Binus/data/videos/VID_20190915_123023.mp4'
PATH_OUT_VIDEO = 'D:/Binus/data/videos/VID_20190915_123023_2.avi'
fo = FaceOrchestrator()
fs = FileVideoStream(PATH_VIDEO).start()
time.sleep(1.0)
frame = fs.read()
h, w, __ = frame.shape
out = cv2.VideoWriter(PATH_OUT_VIDEO,
cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'), 24,
(w, h))
i = 0
while fs.more():
i = i + 1
print('Processing frame ' + str(i))
try:
frame = fs.read()
fo.set_image_to_be_processed(frame)
# detect faces
fo.fd_method = 'AI'
fo.fd_ai_min_score = 0.6
fo.detect_faces()
fo.align_faces(target_face_percent=0.28)
# reconize faces
# fo.fr_max_delta = 0.6
# fo.fr_method = 'DELTA'
fo.fr_min_probability = 0.5
fo.fr_method = 'ML'
frame_with_label, names = fo.recognize_faces()
out.write(frame_with_label)
except:
pass
fs.stop()
out.release()
print('Finish.')
def process_video(file_name, n=1):
"""
iterate over frames of videofile
:param file_name: filename
:param n:process one out of every n frames-> default 1: take all frames
:return:list of finish frames
"""
fvs = FileVideoStream(file_name).start() #load video
cam = cv2.VideoCapture(file_name)
fps = cam.get(cv2.CAP_PROP_FPS) #get original fps of video
counter = 1
frame_list = []
teller = 0
while fvs.running():
if fvs.more():
teller += 1
frame = fvs.read()
if frame is not None:
frame_list.append(
cv2.resize(
frame,
(int(frame.shape[1] / 2), int(frame.shape[0] / 2)))
) #append frame to list and resize it: height/2, width/2
counter += 1
else:
time.sleep(2)
return (frame_list, fps)
def livevideocorrection(filepath, modelname):
fvs = FileVideoStream(filepath, queue_size=128).start()
desired_frames = 250
frame_number = 0
last_time = time.monotonic()
while fvs.more() and frame_number <= desired_frames:
frame = fvs.read()
if frame_number % 10 == 0:
current_time = time.monotonic()
timedif = current_time - last_time
FPS = 10 / timedif
print('FPS:' + str(FPS))
last_time = time.monotonic()
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
prediction = (predict_one(frame,
modelname).squeeze().detach().float().cpu())
prediction = prediction.permute(1, 2, 0).numpy()
prediction = cv2.cvtColor(prediction, cv2.COLOR_BGR2RGB)
cv2.putText(prediction, "Queue Size: {}".format(fvs.Q.qsize()),
(10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
cv2.putText(prediction, ('FPS: ' + str(FPS)), (750, 50),
cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2)
cv2.imshow('preview', prediction)
frame_number += 1
cv2.waitKey(1)
cv2.destroyAllWindows()
fvs.stop()
def main():
# Construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-m", "--model", required=True,
help="path to model file")
ap.add_argument("-c", "--confidence", type=float, default=0.95,
help="minimum probability to filter weak detections")
args = vars(ap.parse_args())
# Initialize our centroid tracker and frame dimensions
ct = CentroidTracker()
# Load our serialized model from disk
print("[INFO] Loading model...")
model = get_model(args["model"])
vs = FileVideoStream("/media/ave/ae722e82-1476-4374-acd8-a8a18ef8ae7a/Rana_vids/Nova_1/121/03-2018.05.25_06.24.23-17.mpg").start()
time.sleep(2.0)
# Loop over the frames from the video stream
while vs.more():
# Read the next frame from the video stream
frame = vs.read()
analyze_frame(frame, ct, model, args["confidence"])
# Cleanup
cv2.destroyAllWindows()
vs.stop()
def UAMS(self):
video_file = FileVideoStream(self.args["video"]).start()
time.sleep(2.0)
fps = FPS().start()
while video_file.more():
# Converting each frame to matrix of numbers
try:
self.frame = video_file.read()
gray = cv2.cvtColor(self.frame, cv2.COLOR_BGR2GRAY)
gray = np.dstack([gray, gray, gray])
except Exception as e:
print(e)
self.frame, self.faces, conf, detection, startX, y = self.cnn_op.CNNOperation(
self.frame, self.net, self.args)
try:
msg = self.calculation(startX, y)
except Exception as e:
print(e)
try:
cv2.imshow("frame-this is main frame", self.frame)
except Exception as e:
print(e)
fps.update()
key = cv2.waitKey(1) & 0xFF
if key == ord("q"):
break
fps.stop()
print("elapsed time : {:.2f}".format(fps.elapsed()))
print("FPS : {:.2f}".format(fps.fps()))
def main(input_video_path, output_video_path, batch_size, async_workers_count):
frame_id = 0
video_path = 'sample_video/car_stopping_shorter_version.mp4'
fvs = FileVideoStream(input_video_path).start()
time.sleep(1.0)
fps = FPS().start()
in_holder_path, op_holder_path = return_io_frames_holder()
print(in_holder_path, op_holder_path)
# loop over frames from the video file stream
input_frames_list = []
while fvs.more():
frame_id += 1
frame = fvs.read()
if (frame is None):
continue
cv2.putText(frame, "Queue Size: {}".format(fvs.Q.qsize()), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
cv2.imwrite(f"{in_holder_path}/{frame_id}.jpeg", frame)
# stop the timer and display FPS information
fps.stop()
print("[INFO] elasped time: {:.2f}".format(fps.elapsed()))
print("[INFO] approx. FPS: {:.2f}".format(fps.fps()))
# do a bit of cleanup
cv2.destroyAllWindows()
fvs.stop()
colonel_batch(in_holder_path, op_holder_path, batch_size,
async_workers_count)
#op_holder_path = "/home/gr8-adakron/Adakron_bay/work/FastAPI/fastapi_exp/video_processing/outputs/bCe7mEE/"
write_detected_video(output_video_path, op_holder_path)
class ImageRouter(QtCore.QObject):
source_is_file = False
on_frame = QtCore.pyqtSignal(np.ndarray)
def __init__(self, filename=''):
super().__init__()
if filename == '':
self.vs = VideoStream(src=0).start()
self.source_is_file = False
else:
self.vs = FileVideoStream(filename).start()
self.source_is_file = True
def __del__(self):
self.vs.stop()
def tick(self):
if self.source_is_file and not self.vs.more():
return False
frame = self.vs.read()
if frame is None:
return False
self.on_frame.emit(frame)
return True
def run(self):
print("[INFO] Started ImageRouter")
while self.tick():
None
print("[INFO] Stopped ImageRouter")
def read_frames(self):
# start the file video stream thread and allow the buffer to
# start to fill
print("[INFO] starting video file thread...")
args = self.argument_parser()
output_path = self.cwd / Path(args.output_dir)
output_path.mkdir(parents=True, exist_ok=True)
fvs = FileVideoStream(args.video).start()
time.sleep(1.0)
# start the FPS timer
fps = FPS().start()
detector = mtcnn.MTCNN()
fno = 0
frames = VideoMeta(args.video).fps()
#print(meta)
# loop over frames from the video file stream
while fvs.more():
fno += 1
print(fno)
# grab the frame from the threaded video file stream, resize
# it, and convert it to grayscale (while still retaining 3
# channels)
frame = fvs.read()
if (fno % frames != 0):
continue
try:
frame = imutils.resize(frame, width=450)
except:
break
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# frame = np.dstack([frame, frame, frame])
if args.detect == "haarcascade":
results = self.face_cascade.detectMultiScale(frame, 1.3, 5)
if len(results) != 0:
self.crop(frame, results[0],
str(output_path / Path(str(fno) + ".jpg")))
else:
results = detector.detect_faces(frame)
# display the size of the queue on the frame
# cv2.putText(frame, "Queue Size: {}".format(fvs.Q.qsize()),
# (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
# show the frame and update the FPS counter
cv2.imshow("Frame", frame)
cv2.waitKey(1)
fps.update()
# stop the timer and display FPS information
fps.stop()
print("[INFO] elasped time: {:.2f}".format(fps.elapsed()))
print("[INFO] approx. FPS: {:.2f}".format(fps.fps()))
# do a bit of cleanup
cv2.destroyAllWindows()
fvs.stop()
def livevideocorrection(filepath, modelname):
fvs = FileVideoStream(filepath, queue_size=128).start()
# ivs = InferenceDataStream(fvs, queue_size=25).start() # --> inference video stream reads from file video stream and forms queue of prepped images for inference
desired_frames = 250
frame_number = 0
last_time = time.monotonic()
while fvs.more() and frame_number <= desired_frames:
frame = fvs.read() # --> frame = ivs.read()
if frame_number % 10 == 0:
current_time = time.monotonic()
timedif = current_time - last_time
FPS = 10 / timedif
print('FPS:' + str(FPS))
last_time = time.monotonic()
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
prediction = (
predict_one(frame, frame_number).squeeze().detach().cpu()
) # --> prediction = predict_one(frame, frame_number) #why? because logic for preparing images is now done asyncronously in ivs.
prediction = prediction.permute(1, 2, 0).numpy()
prediction = cv2.cvtColor(prediction, cv2.COLOR_BGR2RGB)
cv2.putText(prediction, "Queue Size: {}".format(fvs.Q.qsize()),
(10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
cv2.putText(prediction, ('FPS: ' + str(FPS)), (750, 50),
cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2)
cv2.imshow('preview', prediction)
frame_number += 1
cv2.waitKey(1)
cv2.destroyAllWindows()
fvs.stop()
def main(video_path):
video_path = 'sample_video/car_stopping_shorter_version.mp4'
fvs = FileVideoStream(video_path).start()
time.sleep(1.0)
fps = FPS().start()
# loop over frames from the video file stream
input_frames_list = []
number = 0
while fvs.more():
number += 1
frame = fvs.read()
if (frame is None):
continue
cv2.putText(frame, "Queue Size: {}".format(fvs.Q.qsize()), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
ann_frame = stream_frames_to_model(frame, number)
cv2.imshow('frame', ann_frame)
cv2.waitKey(1)
fps.update()
# stop the timer and display FPS information
fps.stop()
print("[INFO] elasped time: {:.2f}".format(fps.elapsed()))
print("[INFO] approx. FPS: {:.2f}".format(fps.fps()))
# do a bit of cleanup
cv2.destroyAllWindows()
fvs.stop()
def main():
vid = (Path(__file__).parent / 'test.mp4').absolute()
stream = FileVideoStream(str(vid), queue_size=100).start()
time.sleep(1)
while stream.more():
img_data = stream.read()
if img_data is None:
break
img = Image(img_data)
transformed = Transformer(img)
image = transformed.cannify()
cropped_img = region(image)
lines = cv2.HoughLinesP(cropped_img.raw,
2,
np.pi / 180,
100,
np.array([]),
minLineLength=40,
maxLineGap=5)
avg_lines = utils.average_slope_intercept(img, lines)
lines_image = img.display_lines(avg_lines)
combined = img.combine(lines_image)
cv2.imshow('', combined.raw)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cv2.destroyAllWindows()
stream.stop()
def standard_run(
self,
x1: int,
x2: int,
y1: int,
y2: int,
sub_ch: list,
channel_len,
):
fps = FPS().start()
cap = FileVideoStream(self.filename).start()
count = 0
start = time.time()
cycle_start = time.time()
frame = cap.read()
frame = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
while cap.more():
# print(count)
frame = cap.read()
if frame is None:
break
if count < 200:
self.frames_buffer.append(frame)
frame = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
frame = frame[y1:y2, x1:x2]
crop = self.mask.apply(frame)
crop = cv2.GaussianBlur(crop, (7, 7), 3.0)
_, crop = cv2.threshold(crop, 150, 255,
cv2.THRESH_BINARY + cv2.THRESH_OTSU)
contours, hierarchy = cv2.findContours(crop, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
for i in range(len(contours)):
avg = np.mean(contours[i], axis=0)
coord = (int(avg[0][0]), int(avg[0][1])) # Coord is (x,y)
ch_pos = int(math.floor((coord[0]) / channel_len))
try:
self.sum_ch1[ch_pos] += float(1)
except:
pass
count += 1
fps.update()
fps.stop()
cycle_end = time.time()
self.cycle_count += 1
end = time.time()
detect_benchmark = end - start
print("Number of frames processed: ", count)
print("Time taken for WBC counting:", detect_benchmark)
print("[INFO] Each cycle time taken = %0.5fs" %
((cycle_end - cycle_start) / count))
return fps
def showVideo(app):
fvs = FileVideoStream(app.vidDir).start()
time.sleep(1.0)
while fvs.more():
frame = fvs.read()
frame = cv2.resize(frame, (720, 480))
cv2.imshow("Frame", frame)
cv2.waitKey(1)
def frames():
camera = FileVideoStream()
time.sleep(1.0)
# loop over frames from the video file stream
while camera.more():
frame = camera.read()
yield frame
def read_frames(self):
# start the file video stream thread and allow the buffer to
# start to fill
print("[INFO] starting video file thread...")
args = self.argument_parser()
output_path = self.cwd / Path(args.output_dir)
output_path.mkdir(parents=True, exist_ok=True)
fvs = FileVideoStream(args.video).start()
time.sleep(1.0)
# start the FPS timer
fps = FPS().start()
detector = mtcnn.MTCNN()
fno = 0
#frames = VideoMeta(args.video).fps()
#print(meta)
e = TfPoseEstimator(get_graph_path("mobilenet_thin"),
target_size=(432, 368))
# loop over frames from the video file stream
while fvs.more():
fno += 1
print(fno)
# grab the frame from the threaded video file stream, resize
# it, and convert it to grayscale (while still retaining 3
# channels)
frame = fvs.read()
#if (fno % frames != 0):
# continue
try:
frame = imutils.resize(frame, width=432, height=368)
except:
break
#frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
#frame = np.dstack([frame, frame, frame])
humans = e.inference(frame)
image = TfPoseEstimator.draw_humans(frame, humans, imgcopy=False)
print("humans:", humans)
# display the size of the queue on the frame
# cv2.putText(frame, "Queue Size: {}".format(fvs.Q.qsize()),
# (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
# show the frame and update the FPS counter
cv2.imshow("Frame", image)
cv2.waitKey(1)
fps.update()
# stop the timer and display FPS information
fps.stop()
print("[INFO] elasped time: {:.2f}".format(fps.elapsed()))
print("[INFO] approx. FPS: {:.2f}".format(fps.fps()))
# do a bit of cleanup
cv2.destroyAllWindows()
fvs.stop()
def piscadas():
vs = FileVideoStream(0).start()
fileStream = True
EYE_AR_THRESH = 0.3
EYE_AR_CONSEC_FRAMES = 3
COUNTER = 0
TOTAL = 0
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor('shape_predictor_68_face_landmarks.dat')
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
while True:
if fileStream and not vs.more():
break
frame = vs.read()
frame = imutils.resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
rects = detector(gray, 0)
for rect in rects:
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
leftEAR = get_ear(leftEye)
rightEAR = get_ear(rightEye)
ear = (leftEAR + rightEAR) / 2.0
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
if ear < EYE_AR_THRESH:
COUNTER += 1
else:
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
COUNTER = 0
cv2.putText(frame, "Piscadas: {}".format(TOTAL), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.imshow("Frame", frame)
key = cv2.waitKey(1) & 0xFF
def write_svg_haar(stream_url):
'''
Reads an alternative face detection model, and connects to the in-memory
Redis database. Detects faces (no identification) in the specified stream
and calculates the corresponding bounding boxes. Writes the bounding boxes
for all detected faces to an svg overlay which is then saved to
Redis to be accessed by other processes.
'''
print("[INFO] opening redis connection")
redis_db = redis.StrictRedis(host="localhost", port=6379, db=0)
print("[INFO] starting stream")
# capture = cv2.VideoCapture(stream_url)
# capture.set(cv2.CAP_PROP_BUFFERSIZE, 0)
fvs = FileVideoStream(stream_url, queue_size=1).start()
process_this_frame = True
while True:
# if process_this_frame:
if True:
# read_flag, img = capture.read()
if not fvs.more():
continue
img = fvs.read()
face_cascade = cv2.CascadeClassifier(
"./haarcascade_frontalface_default.xml")
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray, minSize=(20, 20))
svg_document = svgwrite.Drawing(size=("1280px", "720px"))
svg_document.add(
svg_document.rect(insert=(0, 0),
size=("1280px", "720px"),
stroke_width="10",
stroke="green",
fill="rgb(0,0,0)",
fill_opacity=0))
for (x, y, w, h) in faces:
x = int(x)
y = int(y)
svg_document.add(
svg_document.rect(insert=(x, y),
size=("{}px".format(w),
"{}px".format(h)),
stroke_width="10",
stroke="yellow",
fill="rgb(0,0,0)",
fill_opacity=0))
redis_db.set('overlay', svg_document.tostring())
process_this_frame = not process_this_frame
key = cv2.waitKey(1) & 0xFF
# if the `q` key was pressed, break from the loop
if key == ord("q"):
break
def detect_blinks(videoFilePath):
frameCount = 0
blinkCount = 0
faceDetector = dlib.get_frontal_face_detector()
landmarksPredictor = dlib.shape_predictor(shapePredictorPath)
leftEyeIdx = face_utils.FACIAL_LANDMARKS_IDXS['left_eye']
rightEyeIdx = face_utils.FACIAL_LANDMARKS_IDXS['right_eye']
if FILE_VIDEO_STREAM:
vs = FileVideoStream(videoFilePath).start()
else:
vs = VideoStream().start()
time.sleep(0.5)
target_face_save_path = os.path.join(CURRENT_FOLDER_PATH,
str(time.time()) + '.jpg')
try:
while True:
if FILE_VIDEO_STREAM and not vs.more():
break
frame = vs.read()
frame = rotate(frame, 90)
frame = imutils.resize(frame, width=500)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
target_face = gray
faces = faceDetector(gray, 0)
for (i, face) in enumerate(faces):
(x, y, w, h) = face_utils.rect_to_bb(face)
landmarks = landmarksPredictor(gray, face)
landmarks = face_utils.shape_to_np(landmarks)
leftEye = landmarks[leftEyeIdx[0]:leftEyeIdx[1]]
rightEye = landmarks[rightEyeIdx[0]:rightEyeIdx[1]]
leftEAR = calculateEAR(leftEye)
rightEAR = calculateEAR(rightEye)
ear = (leftEAR + rightEAR) / 2.0
if ear < EAR_THRESHOLD:
frameCount += 1
else:
target_face = gray
if frameCount >= EAR_CONSEC_FRAMES:
blinkCount += 1
frameCount = 0
cv2.imwrite(target_face_save_path, target_face)
except:
blinkCount = 0
vs.stop()
return blinkCount, target_face_save_path
def face_detection_in_video(path2video, seconds_per_frame=1):
print(path2video)
detected_faces = []
face_embeddings = []
facecodes = []
print("[INFO] starting video file thread...")
fvs = FileVideoStream(path2video)
org_frame_rate = fvs.stream.get(5)
width = fvs.stream.get(3) # float
height = fvs.stream.get(4) # float
no_of_frames_to_skip = org_frame_rate * seconds_per_frame
no_of_frames_to_skip = no_of_frames_to_skip if no_of_frames_to_skip >= 1 else 1
print("[INFO] original frame rate: {}".format(org_frame_rate))
print("[INFO] no_of_frames_to_skip: {}".format(no_of_frames_to_skip))
fvs.start()
# start the FPS timer
fps = FPS().start()
frame_no = 1
# loop over frames from the video file stream
while fvs.more():
# grab the frame from the threaded video file stream, resize
# it, and convert it to grayscale (while still retaining 3
# channels)
frame = fvs.read()
if frame is not None:
if frame_no % math.floor(no_of_frames_to_skip) == 0:
data = detect_faces_from_frame(frame, frame_no, width, height)
detected_faces.extend(data["faces"])
face_embeddings.extend(data["face_embeddings"])
facecodes.extend(data["facecodes"])
frame_no += 1
# update the FPS counter
fps.update()
# stop the timer and display FPS information
fps.stop()
print("[INFO] elasped time: {:.2f}".format(fps.elapsed()))
print("[INFO] approx. FPS: {:.2f}".format(fps.fps()))
print("[INFO] total faces detected: {} ".format(len(detected_faces)))
print("[INFO] total frames: {} ".format(frame_no))
return {
"faces": detected_faces,
"face_embeddings": face_embeddings,
"facecodes": facecodes
}
def video_extract_features(vid_path, frame_number, end_frame):
# FASTER video reading (decode in separate thread)
fvs = FileVideoStream(vid_path, start_frame=frame_number).start()
time.sleep(1.0)
current_frame = frame_number - 1
frameCount = end_frame - frame_number + 1
roi = [0.25, 0.55, 0.35, 0.65]
frame_width = int(fvs.get_width())
frame_height = int(fvs.get_height())
result_width = round(roi[3] * frame_width) - round(roi[2] * frame_width)
result_height = round(roi[1] * frame_height) - round(roi[0] * frame_height)
buf = np.empty((frameCount, result_height, result_width), np.dtype('uint8'))
buf_canny = np.empty((frameCount, result_height, result_width), np.dtype('uint8'))
hist_buf = np.empty((frameCount, 16))
print_iter = 0
while fvs.more():
print_iter += 1
# Capture frame-by-frame
frame = fvs.read()
current_frame += 1
frame_roi = get_ROI(frame, roi)
harris_result = get_harris_feature(frame_roi)
hist_result = extract_frame_histogram(frame_roi)
# canny_result = get_canny_feature(frame_roi)
buf[current_frame - frame_number] = harris_result
hist_buf[current_frame - frame_number] = hist_result
# buf_canny[current_frame - frame_number] = canny_result
# cv2.imshow('Frame', harris_result)
if divmod(print_iter, 60)[1] == 0:
print(f'Progress: {100*(current_frame-frame_number)/(end_frame-frame_number)}%')
# Press Q on keyboard to exit
if cv2.waitKey(10) & 0xFF == ord('q'):
print(f'current_frame: {current_frame}')
break
if current_frame == end_frame:
fvs.stop()
break
#save numpy matrix of feature frames
np.save('library_match_1_segment_harris.npy', buf)
np.save('library_match_1_segment_histogram.npy', hist_buf)
def process_video(analyzed_videos, reference_digits, time_parsable, ts_box, vdir, video):
print("[*] Processing video {} from {}".format(video, vdir.directory))
if video in analyzed_videos:
print("[*] Video has been processed. Checking if processing is complete...")
last_processed_frame = get_last_processed_frame(video)
print("[*] Last processed frame for this video is: ", last_processed_frame)
else:
last_processed_frame = None
vs = FileVideoStream(os.path.join(vdir.directory, video)).start()
# Reset frame number
f_num = 0
# Allow the buffer some time to fill
time.sleep(2.0)
while vs.more():
frame = vs.read()
f_num += 1
if (last_processed_frame is not None) and (f_num <= last_processed_frame):
print("[*] Current frame is {}. Waiting for {}...".format(f_num, last_processed_frame + 1))
# Give video buffer time to fill so we don't overtake it
time.sleep(0.01)
continue
else:
try:
# Process the timestamp area in the video
frame_time, ts_box = compute_frame_time(frame, reference_digits, time_parsable, ts_box)
except AttributeError:
if frame is None:
print("[!] Frame was none.")
break
print("Something went wrong. Trying again...")
continue
if frame_time is not None:
time_parsable = True
else:
print("[!] Failed to process time, probably because the frame is distorted.")
# Add the frame information to the logging database
add_frame(directory=vdir,
video=video,
time=frame_time,
frame_number=f_num)
print("{} is done being processed. Adding to database of processed videos...".format(video))
add_processed_video(video=video, total_frames=f_num)
vs.stop()
def run_video(path, net):
'''Runs the uploaded video through the juggle counter algorithm
'''
confidence = 0.45
(W, H) = (200, 200)
direction = 1
bounces = 0
prev_cY = 0
frame_num = 0
prev_frame_num = -7
first_detection = True
print("Processing video...")
vs = FileVideoStream(path).start()
while vs.more():
frames_per_read = 2
for i in range(frames_per_read):
frame = vs.read()
frame_num += 1
try:
frame = cv2.resize(frame, (W, H))
except cv2.error as e:
print("...end of video")
break
detection = process_frame(frame, net, W, H)
if detection[0, 0, 0, 2] > confidence:
(cX, cY, bounces, direction, prev_cY, prev_frame_num,
diameter_temp) = check_detection(frame_num, prev_frame_num, frame,
detection, W, H, bounces,
direction, prev_cY)
if first_detection:
diameter = diameter_temp
first_detection = False
cv2.putText(frame, str(bounces), (int(W * 0.86), int(W * 0.2)),
cv2.FONT_HERSHEY_SIMPLEX, int(W * 0.007), (0, 255, 0),
2)
if 0.9 * diameter < diameter_temp < 1.1 * diameter:
cv2.circle(frame, (cX, cY), int(W * 0.03), (0, 255, 0), -1)
cv2.imshow("Frame", frame)
key = cv2.waitKey(1) & 0xFF
# Only for debuggin:
# if the `q` key was pressed, break from the loop
if key == ord("q"):
break
cv2.destroyAllWindows()
vs.stop()
return bounces
def process_video(video, tolerance=0.6, model="hog", upsample=1):
# process video file and produce clustering results
process_frame = True
frames_with_face = 0
known_faces_encodings = list()
data = list()
print("[INFO] Video name: " + video)
print("[INFO] Starting video processing...")
fvs = FileVideoStream(video).start()
time.sleep(1.0)
# division by two because every second frame is processed
total_video_frames = int(fvs.stream.get(cv2.CAP_PROP_FRAME_COUNT)) / 2
fps = FPS().start()
while (fvs.more()):
# read and transform the frame
frame = transform_to_rgb(fvs.read())
if (process_frame):
timestamp = fvs.stream.get(cv2.CAP_PROP_POS_MSEC)
frame_position = fvs.stream.get(cv2.CAP_PROP_POS_FRAMES)
face_locations = fcr.face_locations(frame, upsample, model)
face_encodings = fcr.face_encodings(frame, face_locations)
face_indexes = cluster_faces(known_faces_encodings, face_encodings,
tolerance)
if (len(face_indexes) > 0):
frames_with_face += 1
result = dict(timestamp=timestamp,
frame_position=frame_position,
face_indexes=face_indexes)
data.append(result)
process_frame = not process_frame
# update the FPS counter
fps.update()
fps.stop()
print("[INFO] Processing successfully finished")
print("[INFO] Elasped time: {:.2f}".format(fps.elapsed()))
print("[INFO] Approx. FPS: {:.2f}".format(fps.fps()))
return total_video_frames, frames_with_face, known_faces_encodings, data
def read_video(path_video: str) -> np.ndarray:
assert pathlib.Path(path_video).exists()
print("Reading video")
buf = []
video_stream = FileVideoStream(path_video).start()
count = 0
while video_stream.more():
count += 1
frame = video_stream.read()
if frame is None:
break
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
buf.append(frame)
print(count)
return buf
def realtime(src, modelweightspath):
if (type(src) == str):
cap = FileVideoStream(src).start()
elif (type(src) == int):
cap = WebcamVideoStream(src).start() # for fast fps read for real-time
fps = FPS().start()
#cap = cv2.VideoCapture(0) # for cv2 read
h = 512
w = 384
seg_model = load_bonnet(3, h, w)
seg_model.load_weights(modelweightspath)
check = True
while check:
start = time.time()
#ret,frame = cap.read() # for cv2 read
frame = cap.read() # for fast fps read webcamvideostream
#print(frame.shape) # frame is of: BGR type
frame = cv2.resize(frame, (w, h), interpolation=cv2.INTER_AREA)
cv2.imshow('Input', frame)
print(frame.shape)
IP = []
ip = load_input(frame, h, w)
IP.append(ip)
IP = np.array(IP)
pred = seg_model.predict(IP)
prediction = pred.argmax(axis=-1)
prediction = to_categorical(prediction, 3)
prediction = np.reshape(prediction, (h, w, 3))
prediction = prediction[:, :, [2, 1, 0]]
cv2.imshow('Output', prediction) # 2=> R 0=> B frame => BGR
end = time.time()
fps.update()
print("FrameRate:" + str(1 / (end - start)))
c = cv2.waitKey(1)
if c == 27:
break
if (type(src) == str
): # Check if next frame exists in case of video file
check = cap.more()
fps.stop()
print("AVg FrameRate:" + str(fps.fps()))
cap.stop() # for fast fps read
#cap.release() # for cv2 read
cv2.destroyAllWindows()
def main(service_socket, image_file_name):
#read events from stdin and push results on stdout
# service_socket: in the format tcp://*:port
context = zmq.Context()
socket = context.socket(zmq.REP)
# ZMQ server must be listening on request first
socket.bind(service_socket)
print("[INFO] starting video file thread...")
fvs = FileVideoStream(image_file_name).start()
time.sleep(1.0)
while fvs.more():
frame = fvs.read()
if frame is None:
break
cv2.namedWindow('Video', cv2.WINDOW_NORMAL)
cv2.imshow('Video', frame)
cv2.resizeWindow('Video', 800, 600)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
img = np.array(frame)
out_img = img
# request from the client side
request = socket.recv_json(flags=0)
if request.get('corr_id', False):
result = {
'corr_id': request['corr_id'],
'shape': out_img.shape,
'dtype': str(out_img.dtype)
}
socket.send_json(result, flags=zmq.SNDMORE)
socket.send(out_img, flags=0, copy=False, track=False)
else:
result = {
'corr_id': request['corr_id'],
'shape': None,
'dtype': None
}
socket.send_json(result)
def vplay():
stream = askopenfilename()
ld = AdvancedLaneDetectorWithMemory(opts, ipts, src_pts, dst_pts, 10, 50, 25)
# import the necessary packages
print("[INFO] starting video file thread...")
fvs = FileVideoStream(stream).start()
time.sleep(1.0)
# start the FPS timer
fps = FPS().start()
# loop over frames from the video file stream
while fvs.more():
# grab the frame from the threaded video file stream, resize
# it, and convert it to grayscale (while still retaining 3
# channels)
frame = fvs.read()
frame = imutils.resize(frame,width = 450)
frame = ld.process_image(frame)
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
#frame = np.dstack([frame])
frame = np.dstack([frame, frame, frame])
# display the size of the queue on the frame
# show the frame and update the FPS counter
frame = imutils.resize(frame,width = 900)
cv2.imshow("Frame", frame)
cv2.waitKey(1)
fps.update()
# stop the timer and display FPS information
fps.stop()
print("[INFO] elasped time: {:.2f}".format(fps.elapsed()))
print("[INFO] approx. FPS: {:.2f}".format(fps.fps()))
# do a bit of cleanup
cv2.destroyAllWindows()
fvs.stop()
def num_blinks(video):
vs = FileVideoStream(video).start()
fileStream = True
COUNTER = 0
TOTAL = 0
time.sleep(1.0)
best_frame = None
max_ear = -1
while True:
if fileStream and not vs.more():
break
frame = vs.read()
frame = imutils.resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
rects = detector(gray, 0)
for rect in rects:
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
ear = (leftEAR + rightEAR) / 2.0
print ear
if ear > max_ear:
best_frame = gray
max_ear = ear
if ear < EYE_AR_THRESH:
COUNTER += 1
else:
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
COUNTER = 0
vs.stop()
return TOTAL, best_frame
import sys
import cv2
import time
from imutils.video import FileVideoStream
videoFile = sys.argv[1]
outputDir = sys.argv[2]
fvs = FileVideoStream(videoFile).start()
time.sleep(1.0)
frameIndex = 0
while (fvs.more()):
frame = fvs.read()
cv2.imshow("Frame", frame)
fileName = outputDir + "\\frame_" + str(frameIndex) + ".tiff"
cv2.imwrite(fileName, frame)
frameIndex += 1
cv2.waitKey(1)
cv2.destroyAllWindows()
fvs.stop()
